Change Log for crop_rec_4k #70 [Jenkins]

Back to Project

View Build Information

Mercurial Build Data

Changes since last success

Artifact diff

Embeddable Build Status

Previous Build

Next Build

build #70 : Changes
1. Merged crop_rec_4k into crop_rec_4k_mlv_lite_snd (detail / bitbucket)
2. mlv_snd: made mlv_rec_queue_block optional
  - reduces the execution time of raw_rec_cbr_started in the vsync hook
  - fixes memory leak (mlv_snd expects mlv_rec to free queued blocks) (detail / bitbucket)
3. mlv_snd: made slot count and maximum blocks per slot configurable (detail / bitbucket)
4. mlv_snd: reworked to achieve better sync with mlv_lite:
  - sound buffer allocation is done in raw_rec_cbr_starting (which must now be called after video memory allocation)
  - raw_rec_cbr_started may optionally be called from the vsync hook (which runs from a task with very high priority)
  
  mlv_lite: call raw_rec_cbr_started from the vsync hook, on first frame
  - result: first audio frame is started ~ 2ms after the first video frame is fully captured
  
  mlv_rec: moved raw_rec_cbr_starting after memory allocation
  - required for mlv_snd to complete the audio memory allocation (detail / bitbucket)
5. mlv_snd: rather than trying to queue a WAVI block into the MLV recorder, provide a mlv_fill_wavi
  this way, the MLV recording module can place this block wherever it wants (not necessarily in the queue)
  previous method still kept for compatibility with mlv_rec
  
  mlv_lite: refactored with mlv_fill_wavi and moved raw_rec_cbr_started right before the recording loop (where it should be)
  this should fix audio sync issues when recording starts; not tested (detail / bitbucket)
6. Merged lua_fix into crop_rec_4k
  (timer functions refactor) (detail / bitbucket)
7. Intervalometer: attempt to get sligthly better timing accuracy, to be tested
  http://www.magiclantern.fm/forum/index.php?topic=21617.msg197314#msg197314 (detail / bitbucket)
8. Moved timer functions to timer.h (always included from dryos.h)
  Renamed get_ms_clock_value / get_us_clock_value to get_ms_clock / get_us_clock
  Replaced all hardcoded references to 0xC0242014 with GET_DIGIC_TIMER()
```
find . -not -path '*/\.*' -type f -name '*.c' -o -name '*.h' -exec sed -i 's/get_ms_clock_value/get_ms_clock/g' {} +
find . -not -path '*/\.*' -type f -name '*.c' -o -name '*.h' -exec sed -i 's/get_us_clock_value/get_us_clock/g' {} +
sed -i 's/\*(volatile uint32_t \*)0xC0242014/GET_DIGIC_TIMER()/' */*.c
...
``` (detail / bitbucket)
9. Reworked get_ms_clock_value / get_us_clock_value / get_seconds_clock
  - only update the 64-bit us_clock on each call (overhead from ~ 4us reduced to less than 1us)
  - lower-resolution clocks are always derived from the microsecond clock and computed only on request
  - dropped get_ms_clock_value_fast (detail / bitbucket)
10. imath.h: fix comments for FMT_FIXEDPOINT*; added FMT_FIXEDPOINT6 (detail / bitbucket)
11. Merged 70D_merge_fw112 into crop_rec_4k
  (dummy, except for a minor whitespace fix) (detail / bitbucket)
12. Merged patchmgr into 70D_merge_fw112 (detail / bitbucket)
13. Merged lua_fix into 70D_merge_fw112 (detail / bitbucket)
14. Merged lua_fix into crop_rec_4k (detail / bitbucket)
15. SRM_BUFFER_SIZE for 600D found using QEMU (detail / bitbucket)
16. SRM_BUFFER_SIZE for 60D (detail / bitbucket)
17. SRM_BUFFER_SIZE for several ML cameras (detail / bitbucket)
18. Full-res LV is not available on cameras other than the 5D3 so remove it from the menu if the camera isn't a 5D3. (detail / bitbucket)
19. Merged patchmgr into crop_rec_4k (detail / bitbucket)
20. Fix compiling with CONFIG_DIGIC_POKE (detail / bitbucket)
21. Modules: patch.h no longer needed (detail / bitbucket)
22. Modules: fix enabling "long" modules when using config presets (issue #2789)
  to reproduce: arkanoid.mo -> ML/SETTINGS/MENU.KEY/arkanoid. -> file create error on exfat
  modules with shorter name: adv_int.mo -> ML/SETTINGS/MENU.KEY/adv_int.e -> it appears to work fine, but it's incorrect
  worst case: arkanoid.mo -> ML/SETTINGS/MYPRESET.NAM/adv_i -> it appears to work fine, possible name clash with other modules (detail / bitbucket)
23. shoot: attempt to make public photo capture functions thread safe, to some extent
  a bit of a hack; some serious rework needed to fix this properly
  at least it appears to fix http://www.magiclantern.fm/forum/index.php?topic=21590.0 (detail / bitbucket)
24. Bracketing: debug messages for troubleshooting http://www.magiclantern.fm/forum/index.php?topic=21590.0 (detail / bitbucket)
25. shoot: make the so-called remote_shoot always take pictures, not movies
  (moved movie functionality in lcdsensor/lightsensor - anything else broken?) (detail / bitbucket)
26. lens_take_picture: always wait until the image capture process starts (and optionally wait for it to finish) (detail / bitbucket)
27. Lua: fix camera.wait() not appearing in API docs (detail / bitbucket)
28. selftest: also test get_task_name_from_id, not just get_current_task_name
  (they use different Canon APIs) (detail / bitbucket)
29. Attempt to fix conflict between ML and Canon overlays on recent DIGIC 4 models
  confirmed on 1200D, likely applies to 1300D as well
  http://www.magiclantern.fm/forum/index.php?topic=12627.msg196304#msg196304 (detail / bitbucket)
30. gui-common: remove outdated comment (detail / bitbucket)
31. Re-applied 2642fbe (fix issue #2844) (detail / bitbucket)
32. Patch manager: nicer way to avoid calling _reapply_cache_patches when not present
  (fixes compiling the installer target and removes the need for defining NO_CACHE_PATCHES) (detail / bitbucket)
33. Merged unified into patchmgr (detail / bitbucket)
34. Patch manager: moved reapply_cache_patches into sync_caches
  also made it private (detail / bitbucket)
35. Merged unified into patchmgr (detail / bitbucket)
36. chdk-dng: fix saving uncompressed files after a lossless DNG
  http://www.magiclantern.fm/forum/index.php?topic=19300.msg196806#msg196806 (detail / bitbucket)